Audio recognition fingerpriting from live stream TV

Hello, I need audio recognition FINGERPRITING - ABSOLUTELY SAME WAY LIKE THIS: [login to view URL] Fingerpriting please read: [login to view URL] In database will be thousands of audio files! and need realtime detection - similiar like for second screen ad detection! Task Detection of ad samples/TV broadcast monitoring in a broadcast dump - ads are splitted to 4-5 seconds- i will provide original stram link and audio database with hundreds audio files! Conditions / MUST: Use machine learning 100% accurancy on 4+ second audio fragment live from TV stream, 100% no any NEGATIVE falses (wrong results) = actually are some movie in tv and will give result of cocacola ads for example 1-3% positive falses are permited results will be visible on windows same like MMT live visio, and will be also sended to postgree sql database also with exactly time (coca-cola - 13:09:24 and Date) deadline 20 days + 14 days for TESTING after are project finished ! you AGREE milestone payment will be released only with 100% no any negative falses, 1~3% positive falses are permited(0%positive falses = money bonus for you) ... ( we give you more ~10 days for improve quality - if there negative falses) what is positive, negative false: [login to view URL] [login to view URL]~ich/classes/mumt621_09/fingerprinting/[login to view URL] The false negative rate is generally used to express the robustness. A false negative occurs when the fingerprints of perceptually similar audio clips are too different to lead to a positive match. ... The rate at which this occurs is usually referred to as the false positive rate. Thanks for your helping and cooperation